sy00000
/
mindspore
forked from OSSInnovation/mindspore
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

update document and example of MatrixInverse

This commit is contained in:
liuhe 2021-04-01 16:30:12 +08:00
parent a691eeb645
commit 6df0d60b92
2 changed files with 21 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
mindspore/ops/operations/math_ops.py
View File

@ -4491,29 +4491,33 @@ class MatrixInverse(PrimitiveWithInfer):
adjoint (bool) : An optional bool. Default: False.
Inputs:
- **x** (Tensor) - A matrix to be calculated.
types: float32, double.
- **x** (Tensor) - A matrix to be calculated. The matrix must be at least two dimensions, and the last two
dimensions must be the same size. types: float32, float64.
Outputs:
Tensor, has the same type and shape as input `x`.
Raises:
TypeError: If `adjoint` is not a bool.
TypeError: If dtype of `x` is neither float32 nor double.
TypeError: If dtype of `x` is neither float32 nor float64.
ValueError: If the last two dimensions of `x` is not same size.
ValueError: If the dimension of `x` is less than 2.
Supported Platforms:
``GPU``
Examples:
>>> mindspore.set_seed(1)
>>> x = Tensor(np.random.uniform(-2, 2, (2, 2, 2)), mindspore.float32)
>>> matrix_inverse = P.MatrixInverse(adjoint=False)
>>> x = Tensor(np.array([[[-0.710504 , -1.1207525],
... [-1.7651395 , -1.7576632]],
... [[ 0.52412605, 1.9070215],
... [ 1.3384849 , 1.4274558]]]), mindspore.float32)
>>> matrix_inverse = MatrixInverse(adjoint=False)
>>> output = matrix_inverse(x)
>>> print(output)
[[[-0.39052644 -0.43528939]
[ 0.98761106 -0.16393748]]
[[ 0.52641493 -1.3895369 ]
[-1.0693996 1.2040523 ]]]
[[[ 2.408438 -1.535711 ]
[-2.4190936 0.97356814]]
[[-0.79111797 1.0569006 ]
[ 0.74180895 -0.2904787 ]]]
"""
@prim_attr_register

14
mindspore/ops/operations/nn_ops.py
View File

@ -2755,8 +2755,8 @@ class ApplyRMSProp(PrimitiveWithInfer):
.. math::
\begin{array}{ll} \\
s_{t} = \\rho s_{t-1} + (1 - \\rho)(\\nabla Q_{i}(w))^2 \\
m_{t} = \\beta m_{t-1} + \\frac{\\eta} {\\sqrt{s_{t} + \\epsilon}} \\nabla Q_{i}(w) \\
s_{t} = \rho s_{t-1} + (1 - \rho)(\nabla Q_{i}(w))^2 \\
m_{t} = \beta m_{t-1} + \frac{\eta} {\sqrt{s_{t} + \epsilon}} \nabla Q_{i}(w) \\
w = w - m_{t}
\end{array}
@ -2850,12 +2850,12 @@ class ApplyCenteredRMSProp(PrimitiveWithInfer):
The updating formulas of ApplyCenteredRMSProp algorithm are as follows,
.. math::
/begin{array}{ll} \\
g_{t} = \\rho g_{t-1} + (1 - \\rho)\\nabla Q_{i}(w) \\
s_{t} = \\rho s_{t-1} + (1 - \\rho)(\\nabla Q_{i}(w))^2 \\
m_{t} = \\beta m_{t-1} + \\frac{\\eta} {\\sqrt{s_{t} - g_{t}^2 + \\epsilon}} \\nabla Q_{i}(w) \\
\begin{array}{ll} \\
g_{t} = \rho g_{t-1} + (1 - \rho)\nabla Q_{i}(w) \\
s_{t} = \rho s_{t-1} + (1 - \rho)(\nabla Q_{i}(w))^2 \\
m_{t} = \beta m_{t-1} + \frac{\eta} {\sqrt{s_{t} - g_{t}^2 + \epsilon}} \nabla Q_{i}(w) \\
w = w - m_{t}
/end{array}
\end{array}
where :math:`w` represents `var`, which will be updated.
:math:`g_{t}` represents `mean_gradient`, :math:`g_{t-1}` is the last momentent of :math:`g_{t}`.